Increasing simulation diversity for autonomous vehicles

This invention describes a way to make simulated environments for training self-driving cars more varied. It works by taking an existing simulated scene, picking out objects within it, and then deciding whether to add extra details, like accessories or features, to specific points on those objects. This decision is made based on a probability assigned to each attachment point, and if an attachment is added, it's chosen from a list of compatible options. The claims specifically narrow the invention to this method of adding attachments based on probabilities.